His ongoing project of the Spartan Constitution (attributed to Xenophon) has no background music. The first chapter is out on YouTube; I’m eagerly anticipating the rest.
For extensive recordings that stick to Allen’s recommendations, you almost have to go with Stratakis, or else Daitz at http://www.bolchazy.com/Assets/Bolchazy/ClientPages/iPodius.aspx. Daitz’s prosody is very annoying, but his pronunciation of the segments is conscientious in the extreme, and he recorded a lot of worthwhile literature. Just to get examples of words that illustrate the phonetic descriptions in Vox Graeca, Forvo should be fine. I haven’t used that site enough to be able to recommend particular readers there.
My pronunciations follow a slightly different philosophy from the recommendations in Vox Graeca. I feel the primary value of the book is in assembling and weighing the evidence for how Greek was spoken in the ancient world; the practical recommendations are secondary. In them, Allen’s goal was to assemble a scheme, based on the best available evidence as surveyed in the book, but also suitable for classroom use with English-speaking students. For that reason, he kept to a direct mapping of each written consonant, vowel or diphthong to a single sound as spoken, and chose not to recommend certain readings (such as /øj/ for οι, or the tone accent) that he felt were too difficult, or too far from the traditional British Erasmian pronunciation.
What you hear in my recordings is based on the idea that the new Attic orthography of the late 5th c. BCE was an approximate phonetic transcription of a dialect that had been transmitted orally up to that point. As such, I assume that each letter can have several slightly different pronunciations in different words. I try to make the inflectional morphology as regular as possible in the underlying phonemic representation, and then apply complex, but intuitive, juncture rules that make the word as expressed sound passably close to the written form used to write it. For instance, it just makes sense to me that if two short vowels sound different, they will also sound different when lengthened: you can make finer distinctions in long vowels. Now, when ε and α are augmented, they both become (written) η. For this reason (and others), I preserve an older distinction whereby η represented two different sounds, roughly /ε:/ and /æ:/, that later merged. The evidence suggests that they were already merging, at least in less cultivated speech, in the 5th/4th c. BCE, but I follow the system with the better internal logic, even at the risk of anachronism. (My take on the “Doric” spellings in tragic choruses: I propose that this distinction was observed throughout the performance of a play, but that professional actors could be trusted to know which η’s take which pronunciation, whereas choristers sometimes needed to be reminded.)
I feel that any historically-informed system with a direct mapping of letters to sounds (including the Vox Graeca system and Buth’s Imperial Koine) is best described as “Revised Erasmian”, whereas my approach is true reconstruction. However, my way is a lot more work. For my current Herodotus recording project, I’m looking up many words in Frisk’s Etymological Dictionary to figure out how they should be pronounced in my system; in older recordings, I usually just made educated guesses, sometimes wrong ones. For my Septuagint recordings, I’m using a simpler scheme, roughly halfway between Vox Graeca and Buth. One unusual sound that I’m trying out in both my current projects (Herodotus and LXX) is a voiced pharyngeal, /ɦ/ or /ʕ/, for the rough breathing. It gives an articulation of words that is essentially psilotic, without losing a valuable distinction for the listener. Also, in my older recordings, the intonation focused almost entirely on phrase-level patterns that express sentence structure; in my newer ones, I’m putting more attention to sentence-level patterns expressing rhetorical function, such as distinguishing questions from statements, or continuation pauses from full stops.
)